--- trunk/run.pl 2005/11/25 00:22:58 141 +++ trunk/run.pl 2006/01/22 13:38:17 389 @@ -7,108 +7,226 @@ use Data::Dumper; use lib './lib'; +use WebPAC::Common 0.02; use WebPAC::Lookup; -use WebPAC::Input::ISIS; -use WebPAC::DB 0.02; +use WebPAC::Input 0.03; +use WebPAC::Store 0.03; use WebPAC::Normalize::XML; use WebPAC::Output::TT; -use WebPAC::Output::Estraier; +use WebPAC::Output::Estraier 0.08; use YAML qw/LoadFile/; +use Getopt::Long; +use File::Path; +use Time::HiRes qw/time/; -my $limit = shift @ARGV; +=head1 NAME -my $config = LoadFile('conf/config.yml'); +run.pl - start WebPAC indexing -print "config = ",Dumper($config); +B -my $type = lc($config->{input}->{type}); +Options: -die "I know only how to handle input type isis, not '$type'!\n" unless ($type eq 'isis'); +=over 4 -my $abs_path = abs_path($0); -$abs_path =~ s#/[^/]*$#/#; +=item --offset 42 -my $lookup = new WebPAC::Lookup( - lookup_file => $config->{input}->{lookup}, -); +start loading (all) databases at offset 42 +=item --limit 100 +limit loading to 100 records -my $isis = new WebPAC::Input::ISIS( - code_page => $config->{webpac}->{webpac_encoding}, - limit_mfn => $config->{input}->{limit}, -); +=item --clean -my $maxmfn = $isis->open( - filename => $config->{input}->{path}, - code_page => $config->{input}->{encoding}, # database encoding -); +remove database and Hyper Estraier index before indexing -my $path = './db/'; +=item --one=database_name -my $db = new WebPAC::DB( - path => $config->{webpac}->{db_path}, -); +reindex just single database -my $n = new WebPAC::Normalize::XML( -# filter => { 'foo' => sub { shift } }, - db => $db, - lookup_regex => $lookup->regex, - lookup => $lookup, -); +=item --config conf/config.yml -$n->open( - tag => $config->{normalize}->{tag}, - xml_file => $config->{normalize}->{path}, -); +path to YAML configuration file -my $out = new WebPAC::Output::TT( - include_path => $config->{webpac}->{template_path}, - filters => { foo => sub { shift } }, -); +=back + +=cut + +my $offset; +my $limit; + +my $clean = 0; +my $config = 'conf/config.yml'; +my $debug = 0; +my $one_db_name; -my $est = new WebPAC::Output::Estraier( - %{ $config->{hyperestraier} } +GetOptions( + "limit=i" => \$limit, + "offset=i" => \$offset, + "clean" => \$clean, + "one=s" => \$one_db_name, + "config" => \$config, + "debug" => \$debug, ); +$config = LoadFile($config); + +print "config = ",Dumper($config) if ($debug); + +die "no databases in config file!\n" unless ($config->{databases}); + my $total_rows = 0; +my $start_t = time(); + +while (my ($database, $db_config) = each %{ $config->{databases} }) { + + next if ($one_db_name && $database !~ m/$one_db_name/i); -for ( 0 ... $isis->size ) { + my $log = _new WebPAC::Common()->_get_logger(); - my $row = $isis->fetch || next; + # + # open Hyper Estraier database + # - my $mfn = $row->{'000'}->[0] || die "can't find MFN"; + my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); + $est_config->{database} = $database; + $est_config->{clean} = $clean; - my $ds = $n->data_structure($row); + my $est = new WebPAC::Output::Estraier( %{ $est_config } ); -# print STDERR Dumper($row, $ds); - -# my $html = $out->apply( -# template => 'html_ffzg.tt', -# data => $ds, -# ); -# -# # create test output -# -# my $file = sprintf('out/%02d.html', $mfn ); -# open(my $fh, '>', $file) or die "can't open $file: $!"; -# print $fh $html; -# close($fh); -# -# $html =~ s#\s*[\n\r]+\s*##gs; -# -# print STDERR $html; - - $est->add( - id => $mfn, - ds => $ds, - type => $config->{hyperestraier}->{type}, + # + # now WebPAC::Store + # + my $abs_path = abs_path($0); + $abs_path =~ s#/[^/]*$#/#; + + my $db_path = $config->{webpac}->{db_path} . '/' . $database; + + if ($clean) { + $log->info("creating new database $database in $db_path"); + rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); + } else { + $log->info("working on $database in $db_path"); + } + + my $db = new WebPAC::Store( + path => $db_path, + database => $database, + debug => $debug, ); - $total_rows++; -}; + # + # now, iterate through input formats + # + + my @inputs; + if (ref($db_config->{input}) eq 'ARRAY') { + @inputs = @{ $db_config->{input} }; + } elsif ($db_config->{input}) { + push @inputs, $db_config->{input}; + } else { + $log->info("database $database doesn't have inputs defined"); + } + + my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; + + foreach my $input (@inputs) { + + my $type = lc($input->{type}); + + die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); + + my $lookup = new WebPAC::Lookup( + lookup_file => $input->{lookup}, + ); + + my $input_module = $config->{webpac}->{inputs}->{$type}; + + $log->info("working on input $input->{path} [$input->{type}] using $input_module"); + + my $input_db = new WebPAC::Input( + module => $input_module, + code_page => $config->{webpac}->{webpac_encoding}, + limit => $limit || $input->{limit}, + offset => $offset, + lookup => $lookup, + ); + $log->logdie("can't create input using $input_module") unless ($input); + + my $maxmfn = $input_db->open( + path => $input->{path}, + code_page => $input->{encoding}, # database encoding + ); + + my $n = new WebPAC::Normalize::XML( + # filter => { 'foo' => sub { shift } }, + db => $db, + lookup_regex => $lookup->regex, + lookup => $lookup, + prefix => $input->{name}, + ); + + my $normalize_path = $input->{normalize}->{path}; + + if ($normalize_path =~ m/\.xml$/i) { + $n->open( + tag => $input->{normalize}->{tag}, + xml_file => $input->{normalize}->{path}, + ); + } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { + $n->open_yaml( + path => $normalize_path, + tag => $input->{normalize}->{tag}, + ); + } + + foreach my $pos ( 0 ... $input_db->size ) { + + my $row = $input_db->fetch || next; + + my $mfn = $row->{'000'}->[0]; + + if (! $mfn || $mfn !~ m#^\d+$#) { + $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); + $mfn = $pos; + push @{ $row->{'000'} }, $pos; + } + + my $ds = $n->data_structure($row); + + $est->add( + id => $input->{name} . "/" . $mfn, + ds => $ds, + type => $config->{hyperestraier}->{type}, + ); + + $total_rows++; + } + + }; + + my $dt = time() - $start_t; + $log->info("$total_rows records indexed in " . + sprintf("%.2f sec [%.2f rec/sec]", + $dt, ($total_rows / $dt) + ) + ); + + # + # add Hyper Estraier links to other databases + # + if (ref($db_config->{links}) eq 'ARRAY') { + foreach my $link (@{ $db_config->{links} }) { + $log->info("adding link $database -> $link->{to} [$link->{credit}]"); + $est->add_link( + from => $database, + to => $link->{to}, + credit => $link->{credit}, + ); + } + } -my $log = $lookup->_get_logger; +} -$log->info("$total_rows records indexed");